Ruslan Vasilev Email & Phone Number

London, Gb

• Pre-training YandexGPT models.• Distributed training efficiency: enhancing distributed training algorithms, optimizing GPU-accelerated computation, minimizing communication & memory overhead.• Model architecture and optimization: including scaling, fast inference and long context.

Московская Область, Россия

• Developed core internal LLM frameworks for pre-training, post-training and model inference.• Enhanced the quality and efficiency of Yandex models through comprehensive experiments in training, architecture, and data.• Open-sourced YaLM-100B: https://github.com/yandex/YaLM-100B

Moscow, Russia

Researched low-resource and parameter-efficient fine-tuning of large transformers. Developed methods were incorporated into different Yandex products. The public submission achieved SOTA in single-model setup on Russian SuperGLUE: https://russiansuperglue.com/login/submit_info/1455

Master'S Degree, Data Science

Machine Learning Engineering

Bachelor'S Degree, Mathematics And Computer Science